This file provides an overview of the matlab code and the data and regression files for 
“How Much Does Housing Collateral Constrain Regional Risk Sharing” 
contains the following files

Computation of the model
The computations of the model, spelled in section 2.6, are done in Matlab. All Matlab files must be stored in the same folder for the code to run. 

The main file is compute_regional_collateral.m

It uses two main functions, one to compute the policy functions, fun_compute.m, and one to simulate (in order to compute the growth rate of the average weight \xi_a), fun_simulate.m.

All other files in the directory are auxiliary programs or store variables or results, such as the parameters of the calibration in calibres_be_lv3.mat.

Data

The data are described in detail in the data Appendix C.
The file Metropolitanconsumptiondata_5102.xls is an excel file that contains the raw data for retail sales, disposable income, and number of households from the publication Sales & marketing Management. The data are annual from 1951-2002 and are described in more detail in Appendix C.3.
United States Regional Aggregate Measures
In Section 3 and Table 1, we use aggregate collateral measures to test the risk sharing hypothesis. A detailed data description is in Data Appendix C.1-3.  The Eviews file RED_LVN_eviews_USaggregate.wf1 contains the data and the regressions. 
The real per household consumption and disposable income share variables are still denoted by cshgrxxx and ishgrxxx. The first measure of housing collateral is based on residential wealth and is called myrw. The labor income growth variable is labgrxxx. The second measure is based on residential fixed assets and is called myrw. There are two corresponding measures of collateral scarcity, as defined in the main text, which are myrwtilde and myfatilde.
The regions are 23 (Combined) Metropolitan Statistical Areas, abbreviated as (xxx=)  ATL (Atlanta), BOS (Boston), CHI (Chicago), CIN (Cincinnati), CLE (Cleveland), DAL (Dallas), DEN (Denver), DET (Detroit), HON (Honolulu), HOU (Houston), KAN (Kansas), LOS (Los Angeles), MIA (Miami), MIL (Milwaukee), MIN (Minneapolis), NYO (New York), PHI (Philadelphia), PIT (Pittsburgh), POR (Portland), SDI (san Diego), SFR (San Francisco), SEA (Seattle), STL (St.-Louis)
The regressions are Seemingly Unrelated Regression with regional fixed effects of CSHGR? on ISHGR?, and either ISHGR?*MYRW  (in rows 1 and 3 of Table 1), or ISHGR?*MYFA  (in rows 2 and 4 of Table 1), or LABGR?*MYRW  (in row 5 of Table 1), LABGR?*MYFA  (in row 6 of Table 1), or by instrumental variables (rows 7 and 8).
United States Regional Collateral Measures
In Section 4.2 and Table 4, we use regional collateral measures instead of national collateral measures. A detailed data description is in Data Appendix C.4.  The Eviews file RED_LVN_eviews_USregional.wf1 contains the data and the regressions. 
The real per household consumption and income share variables are still denoted by cshxxx and ishxxx. The first measure of regional collateral is the home ownership rate hoxxx, the second measure is the regional collateral ratio myxxx. 
The regressions are Seemingly Unrelated Regression with regional fixed effects of LOG(CSH?/CSH?(-1)) on LOG(ISH?/ISH?(-1)), and either LOG(ISH?/ISH?(-1))*HO?/100 (row 1 of Table 4) or  LOG(ISH?/ISH?(-1))*MY?  (row 2), or LOG(ISH?/ISH?(-1))*MY? and MY? (row 3) for 1975-2000.
Canadian data
In Section 4.3 and Table 5, we repeat the analysis with Canadian data. A detailed data description is in Data Appendix C.5. The Eviews file RED_LVN_eviews_Canada.wf1 contains the data and the regressions. 
The regions are the 10 Canadian provinces, abbreviated as alb (Alberta), brc (British Columbia), man (Manitoba), nbr (New Brunswick), nfd (Newfoundland and Labrador), nov (Nova scotia), ont (Ontario), pei (Prince Edward Island), que (Quebec), and sas (Saskatchwan). 
The log change in the real per capita consumption share is denoted the variable cshgrxxx, where the xxx is the abbreviation for the Canadian province.  The log change in the real per capita income share is denoted by ishgrxxx. The regional housing collateral ratio is denoted by the variable myxxx. The Canadian housing collateral ratio is the variable mycan.
The regressions are Seemingly Unrelated Regression with regional fixed effects of CSHGR? on ISHGR?, and either ISHGR?*MYCAN  (in row 1 of Table 5), or ISHGR?*MY?  (row 2), or ISHGR?*MY? and MY?  (row 3) for 1982-2003.